Closure model for intermittent turbulent flows

Closure model for intermittent turbulent flows

@X7-9310/81/111811-12 SO2,CQ/O ,nt. J. Heat Moss Trens~r. Vol. 24, No. 11, pp. 1811-1822, 1981 PergamonPress Ltd. Printedin Great Britain CLOSURE ...

1MB Sizes 0 Downloads 9 Views

@X7-9310/81/111811-12 SO2,CQ/O

,nt. J. Heat Moss Trens~r. Vol. 24, No. 11, pp. 1811-1822, 1981

PergamonPress Ltd.

Printedin Great Britain

CLOSURE MODEL FOR INTERMITTENT TURBULENT FLOWS S. [email protected]* and W. KOLLMANN~ von Karman Institute, Rhode.St.-Genese (Received

30 March 1981)

Abstract-A closure model for turbulent shear flows based exclusively on conditional moments and the intermittency factor is developed. The model contains the equations for intermittency factor, the turbulent zone and non-turbulent zone mean velocities and kinetic energy and dissipation rate for the turbulent zone. The model is applied to the prediction of plane jets and boundary layers and gives satisfactory results for intermittency factor and first order moments.

NOMENCLATURE c,, c,, c3, c4, turbulence model constants; CD9CE19 CE2) CE3, Cmlt Gm flux o3 y defined in equation (29); domain; point set defined by equation (3); flux ; interface term, function ; indicator function ; kinetic energy; natural logarithm; normal vector of interface; probability density function ; pressure ; source; interface expression defined in equation (3), source term ; time ; unitary transformation; velocity ; relative progression velocity of interface; Cartesian coordinates ; Laplace operator ; ensemble average. Greek symbols Dirac function ; 6, dissipation rate; :, scalar variable ; value of Q ; 4% diffusivity ; l-9 intermittency factor ; Y, i Taylor scale ; i, macro-scale ; kinematic viscosity ; V, density ; Pt Prandtl number ; 0, vorticity. W,, Subscripts k, t,

kinetic energy; turbulent ;

coordinate directions; intermittency factor; dissipation rate ; scalar. Superscripts interface ; s, unconditional average; E turbulent zone average; f = non-turbulent zone average; * turbulent zone fluctuation ; non-turbulent zone fluctuation ; 0: unconditional fluctuation.



shear flows exhibit an intermittent character in the neighbourhood of free boundaries. This was first established experimentally by Corrsin [l] and since then a large amount of experimental data for various shear flows has been gathered [2-91. The theoretical treatment of intermittently turbulent flows however proved difficult in particular since certain aspects of it lead to the statistics of multi-valued random functions [lo, 111. The first attempts to establish closed equations for the intermittency factor and selected conditional moments are due to Libby [12,13] who based his model however on a guessed transport equation for the intermittency factor. Dopazo [14] showed how to derive the exact equations for intermittency factor and conditional moments. After that several closure models have appeared in the literature [15,16] which will be discussed in Section 2 of this paper. It is well known that in turbulent shear flows of boundary layer type the intermittency factor follows closely a Gaussian distribution and there seems to be little reason to develop a closure model predicting it as TURBULENT

*Current address: Inst. for Thermodynamics, NTH Trondheim, Norway. t Current address : Dept. Mech. Engng UCD, Davies, CA 95616, U.S.A. 1811

solution of a complex system of partial differential equations. However, there are at least two classes of turbulent flows where the intermittent character is of primary importance and where a Gaussian distribution cannot be assumed: transitional flows and flows with fast chemical reactions. The closure model to be developed in this paper should provide a basis for an improved prediction of such Rows but does not directly address them. Finally it should be noted that any closure model of moment equations involves, in some form or other, assumptions concerning the probabi~ty density function (pdf) of fluctuating variables. If the fluctuations show an intermittent character the contribution from one of the zones (“non-turbulent”) may produce a nearly-singular part of the pdf which in turn makes the pdf strongly non-Gaussian and any assumption based on quasi-Gaussian behaviour will fail. Conditioning of the variables allows the removal of such spikes in the pdf and consequently are closure assumptions based on quasi-Gaussianity better suited for conditional variables.



AND I~ER~i~~~~~

The description of the intermittent structure of turbulent flows requires several definitions based on the notion of a detector or discriminator variable. These quantities will be discussed first. Constant density will be assumed unless stated otherwise. 2.1. Discriminating scalars and interfaces Consider a fluctuating scalar cP(x,t) which satisfies instantaneously the transport equation

and which is non-negative. In order to serve as discriminator Q, should satisfy two conditions: (i) Q should be below a small (compared to a turbulent reference value) threshold value in the nonturbulent region of the flow and be well above the threshold in the turbulent zone. (ii) Cpshould grow with the turbulent zone. By stating these conditions it was implicitly assumed that the turbulent “state” of the flow could be defined uniquely. This is by no means straightforward (Lumley and Panofsky, [17]) but for the present purpose the properties random and vertical are used to define the turbulent state. This definition of turbulence leads to the choice of discriminator enstrophy: cD(x,t) 3 w;U&,


or any monotonically non-decreasing function of it, where w: is the fluctuating vorticity component. The measurement of vorticity is however quite difficult and therefore a class of scalars like temperature, colour, density are accepted as discriminators in suitable circumstances. This raises however the question whether both conditions stated above are satisfied for

scalars different from (2). in particular for scalars that are instantaneously strictly conserved is no reason :CI believe that their growth (and hence the growth of the region where (p is above the threshold) will be the same as (2). Therefore we will use (2) in the following a% discriminator. Consider the solution @(x, t) of (1) for given malt. zation of the velocity field c,(x, t) and the boundary conditions for (9. For differentiable cZ and sufl’icientfy smooth boundary conditions the solution tf, will be differentiable and then the set of points F,(t) for which S(x,r;cp)r~(X.t)---~-0,


and which can be approached as limit of points @(p. 1) < 4, is a surface in the flow domain D. If 4, is ;t discriminating scalar and if the threshold d, > 0 is sufficiently small this surface can be regarded as mathematical model for the turbulent-non-turbulent interface. The real interface has a finite thickness which was estimated by Corrsin and Kistler [l] as being of the order of the Kolmogorov microscale. Only for the limit case of the turbulent Reynolds number going to infinity can we expect to encounter sharp (zero thickness) interfaces because the governing (instantaneous) equations allow only for this limit discontinuous (with respect to the discriminating scalar) solutions. The interface as defined above will therefore be contained within the physical interface and follow its movements. 2.2. Conditimal statistics Conditional statistics are a necessary tool for the description of intermittently turbulent flows. Only single point statistics will be considered here, and questions concerning geometrical properties of the interface such as distance from a coordinate plane leading to statistics of multi-~~alued functions (see [IO, 111) will be excluded. The interface will be assumed differentiable with probability one. Then consider the point (x, r) in the flow domain and define the indicator function 1(x, t) as usual ([l, 141) by l(x,t;

4, z



if S(x. t) 2 0 otherwise


where 4 >, 0 is the threshold value of the discriminator 0. The propagation velocity of the interface F&t) is defined by

where x”,is a point of the interface. Then the indicator function [lS] follows, viz:

where V is the relative progression interface Vu, = L’,- z+;

velocity of the (6)

and II, is the normal vector of F+(r) positive into the

Closure model for intermittent turbulent flows region with I = 1. The normal is defined and unique with probability one which does not exclude however the occurrence with zero probability of intersection points with several different values of n,. The two basic rules for manipulating the indicator function [18] then follow :

whether 0 is strictly conserved or not. The equation for the turbulent zone mean velocity follows from the moment equations by multiplication with I, averaging, and applying (AI-AH). The result can be cast in the form

The term P, includes all influences of the interface fluctuations on the conditional mean velocity and consists of


aI - - v;n,qsj at-


which were already applied to derive equation (5). Denoting ensemble averages by angular brackets we can split any fluctuating quantity 4(x, t) in three different ways :

fluctuation ;



where 6 denotes the turbulent zone average defined by 6 = (I+>l(O;


and 4=J+f/P where 6 denotes defined by

the non-turbulent

4 = ((1 - O#J>/(l - (0).



4 = (4) + 4’ where I#’ is the unconditional


For the closure model and in order to gain a better understanding of the terms in equation (17) the equation for the non-turbulent zone mean velocity gfl shall be given too. For the case that the non-turbulent zone has non-zero vorticity but the fluctuations vz have zero vorticity this equation follows as

(12) zone average



The average (I) of the indicator function is the intermittency factor and denoted by y. From [9] and [13] follow various useful relations between conditional variables which are summarized in the appendix. 2.3. Conditional moment equations The exact moment equations required for the development of the closure model will be given without derivation. The intermittency equation and the equations for conditional mean velocity and conditional turbulent kinetic energy can be found in [ 121 and [14], with a discussion of their properties. The equations for the intermittency factor y can be obtained by ensemble averaging equation (5) for the indicator function and the result can be given as

ay [Y(l -Y)(vZ, $+iFaa,=l n ax 1

$,)I + S, (14)

with source term S, s, = (VS(S)).


This term may still include diffusive, productive and destructive contributions and represents the mean entrainment of non-turbulent fluid into the turbulent zone. Furthermore it is possible that S, depends on the choice of the discriminating scalar 0 in particular

and 2 E +vXv~ denotes the kinetic energy of the irrotational fluctuations. The first terms in equations (17) and (19) have opposite signs if a$&$ and dEJJdx, have the same sign as can be expected in shear layers (jets, boundary 1ayers)and then these terms represent a momentum exchange between turbulent and nonturbulent zones. The second term may have the same sign for both zones. In the high shear region of boundary layer flows the spatial derivative of y is proportional to the crossing frequency of the interface and then the second terms in equations (17) and (19) can be interpreted as momentum sources due to the interface motion. Note that the second term in (17) is a momentum sink if the shear layer is a momentum defect flow (boundary layer, wake) and a momentum source if it is a momentum excess flow (jet). Finally, equations (17) and (19) contain various pointstatistical correlations which have the same structure but opposite signs. These can be interpreted [14] as momentum flux through the interface. The condition of zero vorticity for the fluctuations


S. B~c,c;srdv~.and


08 has not been introduced into (19) but leads to the independent CorrsinKistler [ 141 equations : The interface term F, = F,; and IS grven b)

- t.pL$rta)G(S)>.(20) The transport equation for the turbulent zone Reynolds-stress tensor tXj%~can be given in the form

It is worth noting that the redistributive correlation of pressure and rate of strain in (21) does not contribute to the balance of the conditional kinetic energy because the divergence

+q$yj-ⅈ;,. The symbol t?XB denotes the turbulent pation rate tensor defined by

but is non-fluctuating.

Similarly follows the equation

(21) for the trace Eof the dissipation rate tensor using the zone dissi-

relations given in the Appendix. It can be cast into the form


(22) and F,, represents the collected interface terms :

and the interface term is given by


+ ;

~u:vpvS)~ + ((ofnil + u;n,)p*G(S))

(23) with S;. given by (15). Contraction the equation for the conditional turbulence r

of indices leads to kinetic energy of

(24) (28) The interface terms F, and F‘, have the same structure as (26) and (28) show. The first term is as for


Closure model for intermittent turbulent flows

approach zero as y + 0 whereas [ 31 117~ - o",l+ 0 as (16) a consequence of the introduction of the unconditional mean velocity in the convective part in (25) y + 1, it follows that D, should approach zero faster for y + 1 than for y + 0. Hence for the crossflowand (27). The second term is the product of the component of D, the following are suggested: turbulent flux of Ear Erespectively, times the gradient of the intermittency factor. The third term represents a reduction of I? and z due to the entrainment of nonturbulent fluid. The last term group contains the and production/destruction of E and Zat the interface and the transport through it. 3. CLOSURE MODEL

A closure model on the basis of the k - c model initiated by Jones and Launder [19] will be developed for incompressible flows. The complete model will consist of a set of modelled equations for the intermittency factor plus turbulent zone mean velocity, kinetic energy, dissipation rate and non-turbulent zone mean velocity. Unconditional qualities such as mean velocity can be computed with the formulae given in the appendix. 3.1. Intermittency equation The intermittency equation (14) contains two terms that require closure assumptions. The first is the transport of y due to the relative motion of turbulent and non-turbulent zones: D, = -y(l

- y,(ua - v’,,.


This term could be considered closed if the crossflow component of the mean velocities was included in the model. For boundary layer type flows however the determination of these components becomes very inaccurate and therefore a model expression for this term is proposed. For boundary layer type shear flows the expression

holds, where y is the crossflow coordinate. The coefficient of this derivative should then have the dimension of a (turbulent zone) viscosity and thus

with F(y) still to be determined. This expression would however contradict the experimental evidence in parabolic flows [3, 7, 81 where Ii - I?[ > Ii - v’(.The reason for it is that an important contribution to the flux of y due to the inhomogeneity of the mean velocity has not been included (see Lumley [ 151). The complete model should be:

For complex turbulent flows however the equations for all mean velocity components will be solved and then (30) can be avoided. The modelling of the intermittency source S, requires more detailed considerations. Several models have been proposed so far. Libby [12] proposed the expression 8 ~ ;(

!!El 3:4(1 _y)!?, A u2 >

where U is a reference velocity and A is a length scale of the order of the width of the shear flow. Chevray and Tutu [9] suggested a modification of Libby’s model S, 2 2C,y(l

- y,,


with C, = 3.447, and the length scale A = rliZ the half-width of the round free jet which they studied experimentally. In the analysis of the conditional pdf of a passive scalar I’ in a non-buoyant turbulent plume O’Brien [16] suggested the model S,r

c;Y(r)P(o+) 1s

where D is the molecular diffusivity of I, 1, is a turbulent micro-length scale for I, and P(O+) denotes the limit from I > 0 of the turbulent zone pdf P of the r-fluctuations. Libby’s model produced good agreement between calculated and measured intermittency factor and conditional mean velocity. It was applied however only to self-similar flows and unconditional flow variables were prescribed [12]. Furthermore is this model neither Gahlei-invariant nor applicable to complex flows where 0 might change sign? O’Brien’s model is of limited use also, since the value of P(r) approaching zero from r > 0 might not be bounded as the example beta-function p-1(1

p(r) =



_ r)P-1

drra-‘(1- I-y-1


shows, which leads to P(O+)-+ zofor 0 < a < 1 without an atomic contribution to the unconditional For parabolic flows the velocity difference in longipdf at I- = 0. Furthermore, it would require modelling tudinal direction is not required and for the crossflow of P(O+)in a set of equations not containing the pdf. component the second part can be neglected. ThereThe model suggested here consists of three fore, only F,(y) has to be determined. The limits of (Z= contributions : - v’,)for y -+ 0 and y + 1 indicate what F,(y) should s, z sy + SP) - sy’ be. From the fact that 1~; - cZl does not necessarily (31) , ’



far downstream and for the calculations of these flows A was set zero. The final result for the intermittency equation is then for parabolic plane flows

and each contribution will be discussed in turn. The first part of S., is

and is based on the following considerations. The production of y should be proportional to the production of kinetic energy of turbulence because an increase of k” implies an increase of vorticity fluctuations and therefore a production of the discriminator (2). This production of 7 will however depend on y itself, because the term (32) must not violate the realizability condition for 7, viz. 0 < ;I < 1. This leads to F&r = C ~(1 - v) in lst-order. For turbulence models based on the turbulent viscosity concept the resulting form is for parabolic flows (x,-dominant flow direction)

The second part of S;. is:

The production of y should be proportional to the gradient of 7. In fully developed shear flow turbulence a rapid variation of 7 implies slow movement of turbulent fluid in the direction of the gradient of 7 but rapid orthogonal to it, hence strong anisotropy. Since anisotropy of the Reynolds-stresses is counteracted by a part of the pressureestrain rate correlations this process must be accounted for in the y-equation too and is represented by (34). This term can be interpreted as a non-linear convection which implies that (34) will not violate the realizability condition for above y as long as y is at least twice continuously differentiable. Hence reduces F{f: in lowest order to a constant. The third term St3j is: $3'=C I)


F(3)_ 3 (Y) IF

+ c

-- C,i’(l - 7): t


The model constants are summarized in Table 1 3.2. Conditional moment equatkms The closure of the turbulent zone momentum equation (16) and the equations for the turbulent zone kinetic energy (25) and its dissipation rate (27) is based on the concept of turbulent viscosity as applied by Jones and Launder to the k-c model [19]. In fact the same closure assumptions will be employed but for turbulent zone moments only. For the momentum equation (16), in the dominant flow direction Y two closure assumptions are required. The turbulent zone shear stress is expressed as

analogous to the unconditional model (but not a consequence of it). The second term F, in (16) however does not have a counterpart in the unconditional equation and the last part of (17) requires therefore a new closure assumption. The interface terms in ( 17) are modelled by

and with (30). (37) the model for ( 17) after rearrangement is then


Consider a positive threshold value y of the discriminator defining the interface (3) and a homogeneous turbulence which is not maintained by external forces or boundary conditions. If initially 7 = 1 then the equation for ;’ is

Realizability requires Fc3) to be proportional to 7 and in the limit 4 + 0 to be proportional to 1 - ; because @ = 0 would not be reached within a finite time interval. But for non-zero threshold values the discrimination will slip under the threshold within a finite time interval. Therefore the function F&’ is qT,)=y(l --?+A), where A depends on 4. For developing shear layers and non-homogeneous flows however the influence of A becomes negligible except

The form (38) suggested for the interface terms IS based on the assumption that they are dominated by the mean strain rate in the turbulent zone and the gradient of the intermittency factor as a measure for the crossing frequency. The resulting model (39) acts in shear layers as a momentum sink if the non-turbulent flow is faster and as source if the turbulent flow is faster and therefore increases the spreading rate of the turbulent zone. The resulting closed equation for the turbulent zone mean velocity u”is then

The equations for the turbulent zone kinetic energy R and its dissipation rate require one additional


Closure model for intermittent turbulent flows

be estimated. The experimental evidence in parabolic flows [3,7, 81 shows that the stress tensor cannot be neglected immediately because the normal stress components in the non-turbulent zone can amount to 40% of their turbulent zone counterparts. The shear stress in the non-turbulent zone however is much smaller C,,Y(l - Y)c’ than the turbulent zone shear stress and could be neglected. From this follows that no simple relation between the stress tensors in turbulent and nonturbulent zone can hold component-wise but possibly in more general tensorial form via the unitary transform T,, corresponding to a rotation of the local coordinate system -Y)cs (41) Y 0~~~2 const. f(Y) TZyZ$7& (46)

modelling assumption. The model for F, follows from (30) (37) and the expression (k*v6(S))

+ (upping)

- v

c> n,~W) 1



(k*n,G(S)) 2


for the interface terms as


where S, is given by (31)-(35). In the same manner the model for F, is obtained as

With these two expressions the closed equations for the turbulent zone kinetic energy and its dissipation rate follow as:

;+ F,



The turbulent zone pseudo-viscosity is then analogously to the unconditional case defined by I? y3 = CDT E


The closed equations for turbulent zone mean velocity (40), kinetic energy (43) and dissipation rate (44) reduce for Y approaching unity ro the unconditional model if the first and second derivatives approach zero too. Experiments show that this can indeed be expected for fully developed shear flows [3, 71, in which the Y-profile reaches a nearly-constant (unity) region in the centre part of the flow. The turbulent zone quantities however are not sufficient to determine unconditional moments as (Al) and (A7) prove. Therefore, the closure of equation (18) for the non-turbulent zone velocity is required for the calculation of the unconditional mean velocity. For the closure of equation (18) the stress tensor in the non-turbulent zone and the interface terms F, must

The unitary transform T,, is determined by the angles between the principal axis of the stress tensors. For plane and axisymmetric flows the principal axis of the stress tensor in the non-turbulent zone is close to the gradient of the intermittency factor as the experiments show [7, 81. Hence is this angle given by the principal axis of 0% and the gradient of Y. For the kinetic energies, which are invariants of the two tensors, however follows the simpler relation t=

c, f(Y)R


only requiring the knowledge of the constant and the function f(Y). The data for boundary layers [7, 81 suggest for the constant a value of C, g 0.5 and for f(Y) = Y. The modelling of equation (18) however follows from the fact that the shear stress in the non-turbulent zone is an order of magnitude of smaller than the shear stress in the turbulent zone that the corresponding term can be neglected. This leaves the interface term (19) to be modelled. This term cannot be negligible because following from 3, 2 0 together with z 2 0 is the existence of a laminar flow in the non-turbulent zone at Renumbers well in the turbulent regime which cannot be stable. Considering the order of magnitude of the terms constituting F, the first two are negligible in contrast to (17) where the stress tensor has to be kept. The approximation of an inviscid flow is reasonable for the non-turbulent zone in most cases (except transitional flows), thus the viscous terms in (19) are negligible and F, reduces to PZ


Y (([email protected] + PO+w)>.

Before suggesting any closure expression for this term the boundary conditions for the non-turbulent zone velocity must be considered. For fully developed turbulent shear flows the probability for a patch of non-turbulent fluid to reach the opposite boundary becomes very small and therefore the limit Y + 1 for non-turbulent zone quantities loses its meaning. However these quantities must still remain bounded for ~1 if a boundary value problem for non-mrbulent


S. BYGGST~)YI_and W. Table 1. Model constants

Cl C, C, C, CD

0.2 0.6 0.1 0.5 0.09

(r Ok FL,

1.0 1.0 1.3 0.7

C,, CZL C iz C,,

1.44 I .95 1.5 2.5

The values for C, Us,a,, C,,, C,, are standard values taken from the original k-i: model [l9]. zone variables is to be solved. On the rare occasion of a non-turbulent patch of fluid penetrating to the other parts of the boundary it must satisfy the same conditions as the turbulent eddies. Hence the boundary conditions for the non-turbulent zone velocity should be the same as for the turbulent zone quantity To satisfy this condition the interface term must therefore diffuse the momentum from the boundaries into the flow field and create/destroy momentum there. Recalling the alternative way of representing the Dirac function defined on the surface .7(x, t) = 0 as introduced by Dopazo [ 141 we can write for the first part of F,:

(where uh is a sphere of radius 6 around x and r(u,) its volume) which can now be recognized as momentum flux across the interface per unit volume due to the fluctuations of the relative progression velocity b’ and represents therefore the divergence of a flux. Hence the model


and the boundary layer on a smooth wall at zero pressure gradient are considered for comparison of calculation and experiment. 4.1. Plurze !er The measurements of Gutmark and Wygnanskl I~-i1 in the plane jet provide an excellent basis for comparison with the calculations. The boundary CMditions for this case are zero gradient conditions at the symmetry asis and Dirichlet conditions at the outer edge of the flow domain. At the jet pipe exit rectangular velocity profiles 17 = u’ and a small turbulence level were prescribed and the intermittency factor was set ti> factor ;‘(I == 0.001 at this section. The intermittency develops downstream first in the plane shear layer until the core region of the jet has vanished and thep adjusts to the profile to be expected in the jet as shown in Fig. I. In the self-similar region (Fig. 2) !I:? intermittency factor agrees well with the experimental distribution. The bars in Fig. 2 give the variation of the experimental results if the measurements of Heskestad [5] and Bradbury [4] are taken into account. For ;tn ideal Gaussian distribution of ;‘(x) and no folding ol the interface. the cross-derivative of ,’ would (properi! normalized) be the Gaussian probability density function of the interface position. The calculated i+;:i\ deviates from the measured crossing frequency LXXsiderably and shows that the calculated maximum I\ maximum, closed ttr , =-. 0.5 than the measured Indicating a stronger deviation from Gausslanity in the measurements. The unconditional and turbulent ione mean velocities in Fig. 3 show close agreement IX tween measurement and calculation. The cwreni


is suggested. Since non-turbulent zone fluctuations decay rapidly with distance from the interface [20] the flux ,PX8 will be proportional to ;’ and thereforef(7) = ;‘. The second part of F, containing the pressure fluctuations is neglected because p” is determined by large scale fluctuations whereas PI, corresponds to gradient fluctuations which are dependent upon small scale fluctuations whose correlations be expected to be weak. The closed equation for L?follows now as

The constants

are summarized

in Table 1.


The modelled transport equations for the intermittency factor and the turbulent zone and non-turbulent zone moments are discretized and solved using standard finite-difference procedures [21]. The plane jet



FIG,.I. lmtial development ofintermittency factor m planejst over the range 0 < \-.I) < 10.


Closure model for intermittent turbulent flows

prediction model employing the concept of turbulent viscosity cannot be expected to give excellent results for 2nd-order moments. The fluctuation intensity of velocity in the turbulent zone and unconditional (Fig. 4) normalized with the respective axial values shows the correct distribution across the jet but their difference is overpredicted in the outer part. 06-

4.2. Boundary layer For the calculation of turbulent boundary layers low Re number corrections must be introduced in order to represent properly the region close to the wall where the turbulent Re number drops to zero. The correlations suggested by Jones and Launder [ 191 are applied to the turbulent zone variables without modification. The boundary conditions used in the calculations are Dirichlet conditions flow except for the intermittency

FIG. 2. Intermittency factor y and dy/dy in self-similar region of plane jet compared with measurements [3].


FIG 3. Turbulent zone (solid line) and unconditional (broken line) mean velocities in plane jet compared with experiments [3] (symbols).









at both edges of the

factor which satisfies a zero gradient condition at the wall. The initial profiles are given by r = 0.001 and $x0, y) as laminar boundary layer and u”(x,, y) as turbulent boundary layer profile. The initial boundary layer thicknesses are S(k) = 0.003m and S(C) = 1.76($ and the initial profiles for Land c”correspond to a turbulent boundary layer. The unconditional distributions of velocity and kinetic energy correspond however at the beginning to a laminar boundary layer. The downstream development is dominated by a rapid growth of y in the lower part of the boundary layer which in turn leads to a growth of the unconditional kinetic energy and a change of velocity profile to the turbulent form which is achieved over a distance of the same order of magnitude (x - x0 - 0.4 m) as the transition region. After this initial region the development of the boundary layer is calculated until the Re,-number of the experiments of Kovasznay, Kibens and Blackwelder [7] is reached (6 = 0.09 m). The results at this point are presented in Figs. 5-8. The intermittency factor agrees well with the measurements [7,8] and the derivative dy/dy is much closer to the measured crossing frequency than in the jet (Fig. 2). The fluctuation intensities (Fig. 6) for turbulent zone, unconditional and non-turbulent zone velocity fluctuations agree reasonably with the measured u’/ii, profile. As for the jet, the turbulent zone fluctuations are larger than the measured values. The mean velocities in the outer part of the boundary layer (Fig. 7) show for all three statistics the same form as the experiment. Finally the model (47) for the non-turbulent zone fluctuations of u, is compared in Fig. 8 with the mesrements. According to Philips’ hypothesis [22] (p/u”,)-1” should vary linearly withy. The predictions are slightly curved but fairly close to the experimental values and can be considered satisfactory within the framework of a closure model based on the concept of a turbulent viscosity.

YI 2

FIG.4. Intensities of velocity fluctuations in plane jet (turbulent zone, solid line; unconditional, broken line) compared with experiments [3].



A closure model for turbulent shear flows based exclusively on conditional moments and the intermit-



FIG 5. Intermittency factor 1’and &;idy in turbulent boundary layer (Re, = 21,000) compared with measurements and A [7]), + and x [8].

0 08


FIG. 8. Non-turbulent zone fluctuations measurements [7] and Philips’ hypothesis

r 4

compared wrth (broken line).




I 0






FIG;. 6. Intensity of velocity fluctuations: (1) turbulent (2) unconditional, (3) non-turbulent zone, compared measurements [73 of uI/u ,

zone, with

tency factor was developed. The model contains the equations for intermittency factor, the turbulent zone and non-turbulent zone mean velocities and kinetic energy and dissipation rate for the turb?Jlent zone. The main part of the paper was devoted to the closure of the equation for the intermittency factor: production of ;’ via the kinetic energy created in the turbulent zone and the inhomogeneity of y-field, destruction by viscous effects and turbulent diffusion. The conditional moment equations contain a new-term group representing the effect of moving interface, and models representing them were suggested. The results obtained with this model give satisfactory agreement for intermittency factor and conditional moments of first order. This shows the possibility of treating turbulent flows, in which transport processes across interfaces play an important role (such as turbulent flames) with conditional turbulence models. Such models would allow the transport process across the interface to be dealt with explicitly because there are terms in the equations accounting for them. REFERENCES 1. S. Corrsin and A. L. Kistler, The free-stream


NACA TN 3133 (1954). 2. H. Fiedler and M. R. Head. Intermittency measurements of turbulent

3. 4. 5.

6. FK;. 7. Mean velocities in outer part of boundary layer: (1) turbulent zone, (2) unconditional (3) non-turbulent zone compared with measurements [7].



in the turbulent boundary layer. J. Fluid Me&. 25, 719 (1966). E. Gutmark and I. Wygnanski, The planar turbulent jet, J. Fluid Me&. 73, 465 (1976). L. J. S. Bradbury, The structure of a self-preserving turbulent plane jet, J. Fhid Mech. 23, 31 (1975). G. Heskestad, Hot-wire measurements in a plane turbulent jet, Trans. ASME, Ser. E. J. appl. Mech. 32, 721 (1965). S. T. Paizis and W. H. Schwarz. An investigation of the topography and motion of the turbulent interface, J Fluid Mech. 63, 315 (1974). L. S. G. Kovasznay, V. Kibens and R. F. Blackwelder.


Closure model for intermittent turbulent flows Large-scale motion in the intermittent region of a turbulent boundary layer, J. Fluid Me& 41, 283 (1970). 8. T. B. Hedley and J. F. Keffer, Some turbulent/nonturbulent properties of the outer intermittent region of a boundary layer. .____ J. Fluid Me&. 64, _ 645 (1974). _ -. 9. R. Chevray and N. K. ‘l‘utu, lntermlttency and preferential transport of heat in a round jet, J. Fluid Mech. 88, 133 (1978). 10. S. Corrsin and 0. M. Philips, Contour length and surface area of multiple-valued random variables, J. Sot. indust. uppl. Math. 9, 395 (1961). 11. J. L. Lumley, On multiple-valued random functions, J. math. Phys. 5, 1198 (1964). 12. P. A. Libby, On the prediction of intermittent turbulent flows, J. Fluid Mech. 68, 273 (1975). 13. P. A. Libby, Prediction of the intermittent turbulent wake of a heated cylinder, Phys. Fluids 19, 4941 (1976). 14. C. Dopazo, On conditional averages for intermittent turbulent flows, J. Fluid Mech. 81,433 (1977). 15. J. L. Lumley, Second order modelling of turbulent flows, in Prediction Methodsfor Turbulent Flows (Edited by W. Kollmann), Vol. 1, Hemisphere (1980). 16. E. E. O’Brien, Stochastic properties of scalar quantities advected by a non-buoyant plume, J. Fluid Mech. 89,

This relation shows that the unconditional correlation is not necessarily between the conditional correlations but may rc*rv be larger or smaller than &*$* and #‘$O. The differentiation rules (7) and (8) for the indicator function and (9)-(13) imply the following relations: h,

-d4’ = - f ax,



and rv (A9) Note that (AS) and (A9) imply that the point-statistical mean values of differentiated fluctuations do not necessarily vanish. Finally the correlations involving differentiation can be evaluated as


209 (1978).

17. J. L. Lumley and H. A. Panofsky, The structure of atmospheric turbulence. Interscience (1964). 18. I. M. Gelfand and G. E. Shilov, Generalizedfunctions. Vol. 1. Academic Press (1964). 19. W. P. Jones and B. E. Launder, The prediction of laminarization with a two equation model of turbulence, lnt. J. Heat Mass Transjkr 15, 2, 301 (1972). 20. A. A. Townsend, The structure of turbulent shear jiows.

(2nd ed.) Cambridge University Press (1976). 21. S. V. Patankar and D. B. Spalding, Heat and mass transfer in boundary layers. Intertext, London (1970). 22. 0. M. Philips, The entrainment interface, J. Fluid Mech. 51, 97 (1972).


For non-turbulent zone quantities follows h,

w” -=

ax, &







and for correlations rJ h*

APPENDIX The conditional statistics introduced in Section 2.2 imply

various relations between conditional and unconditional quantities. The mean values are related by

as a consequence of (11) and (13). The proper mean values of the fluctuating parts are &&O, but mean values with respect to different generally non-zero. In particular the following for mean values of fluctuating components. ditional fluctuations averaged unconditionally

statistics are relations hold For the conwe obtain :

= (1 - Y)($ - $),


<9O> = Y(i - &)*


and for the unconditional fluctuation it follows that:

8 = - <4*>,


4 =-



Specialization of r$ and $ can be used to produce relations for higher order moments. Finally we note a consequence of (A6) and the fact that the velocity is continuous across the interface for the point-statistical part of the momentum sources F,. Denote by

S, = <(W

+ pn,)W))

- v(~“p(S1)

Furthermore we note that - v& @ = y4* + (1 - Y)@.


YG+(1 -

s,+y+so(1-Y)=s’+2(B_v”)dydy # il ” 2

y)G + Y(1 - Y)($ - 4)($ - $).


The Znd-order moments are given by




from (A6) and (A7).

= ax,ax,





R&m&On dtveloppe un modkle de fermeture pour les tcoulements turbulents “Shear flows”, exclusivement Bpartir des moments conditionnels et du facteur d’intermittence. Le modble contient les tyuations du facteur


cinktique plans

des vitesses

et du taux

de dissipation

et des couches


moyennes pour

et ii donne

de la zone

la zone

des rtsultats moments





du premier

et de la zone mod&le pour


est appliquk le facteur

de I’Cnergie

g la prkvision


des jets

et pour



Zusammenfassung-Ein Schliessungsmodell fiir turbulente Scherstrtimungen, das nur auf bedingten Momenten und Intermittenzfaktor beruht, wird entwickelt. Das Model1 umfasst die Gleichumgen fiir Intermittenzfaktor, mittlere Geschwindigkeiten in turbulenter und nichtturbulenter Zone sowie kinetische Energie und Dissipationsrate in der turbulenten Zone. Das Model1 zeigt bei Anwendung auf den ebenen Freistrahl und die Grenzschicht zufriedenstellende Resultate fiir Intermittenzfaktor und Momente erster Ordung.



MOAE.JIb flJI5I [email protected]




a TaKYe



CpeZlHHX 3Hepl-kfH

AcnonbsyeTcn nna pacseTa pe3yJIbTaTbI



W [email protected]@iUBeHTe

nepMemaeMOCTii, KHHeTU’feCKOii






CnBHrOBOrO Monenb





CKOpOCTeii B 30HaX Typ6yJieHTHOrO

W CKOpOCTki cTpyii


5i norpamiwibrx


Ll 30He



cnoen w naer